gdb/bcache.h

   1 /* Include file cached obstack implementation.
   2    Written by Fred Fish <fnf@cygnus.com>
   3    Rewritten by Jim Blandy <jimb@cygnus.com>
   4
   5    Copyright (C) 1999-2000, 2002-2003, 2007-2012 Free Software
   6    Foundation, Inc.
   7
   8    This file is part of GDB.
   9
  10    This program is free software; you can redistribute it and/or modify
  11    it under the terms of the GNU General Public License as published by
  12    the Free Software Foundation; either version 3 of the License, or
  13    (at your option) any later version.
  14
  15    This program is distributed in the hope that it will be useful,
  16    but WITHOUT ANY WARRANTY; without even the implied warranty of
  17    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  18    GNU General Public License for more details.
  19
  20    You should have received a copy of the GNU General Public License
  21    along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
  22
  23 #ifndef BCACHE_H
  24 #define BCACHE_H 1
  25
  26 /* A bcache is a data structure for factoring out duplication in
  27    read-only structures.  You give the bcache some string of bytes S.
  28    If the bcache already contains a copy of S, it hands you back a
  29    pointer to its copy.  Otherwise, it makes a fresh copy of S, and
  30    hands you back a pointer to that.  In either case, you can throw
  31    away your copy of S, and use the bcache's.
  32
  33    The "strings" in question are arbitrary strings of bytes --- they
  34    can contain zero bytes.  You pass in the length explicitly when you
  35    call the bcache function.
  36
  37    This means that you can put ordinary C objects in a bcache.
  38    However, if you do this, remember that structs can contain `holes'
  39    between members, added for alignment.  These bytes usually contain
  40    garbage.  If you try to bcache two objects which are identical from
  41    your code's point of view, but have different garbage values in the
  42    structure's holes, then the bcache will treat them as separate
  43    strings, and you won't get the nice elimination of duplicates you
  44    were hoping for.  So, remember to memset your structures full of
  45    zeros before bcaching them!
  46
  47    You shouldn't modify the strings you get from a bcache, because:
  48
  49    - You don't necessarily know who you're sharing space with.  If I
  50    stick eight bytes of text in a bcache, and then stick an eight-byte
  51    structure in the same bcache, there's no guarantee those two
  52    objects don't actually comprise the same sequence of bytes.  If
  53    they happen to, the bcache will use a single byte string for both
  54    of them.  Then, modifying the structure will change the string.  In
  55    bizarre ways.
  56
  57    - Even if you know for some other reason that all that's okay,
  58    there's another problem.  A bcache stores all its strings in a hash
  59    table.  If you modify a string's contents, you will probably change
  60    its hash value.  This means that the modified string is now in the
  61    wrong place in the hash table, and future bcache probes will never
  62    find it.  So by mutating a string, you give up any chance of
  63    sharing its space with future duplicates.
  64
  65
  66    Size of bcache VS hashtab:
  67
  68    For bcache, the most critical cost is size (or more exactly the
  69    overhead added by the bcache).  It turns out that the bcache is
  70    remarkably efficient.
  71
  72    Assuming a 32-bit system (the hash table slots are 4 bytes),
  73    ignoring alignment, and limit strings to 255 bytes (1 byte length)
  74    we get ...
  75
  76    bcache: This uses a separate linked list to track the hash chain.
  77    The numbers show roughly 100% occupancy of the hash table and an
  78    average chain length of 4.  Spreading the slot cost over the 4
  79    chain elements:
  80
  81    4 (slot) / 4 (chain length) + 1 (length) + 4 (chain) = 6 bytes
  82
  83    hashtab: This uses a more traditional re-hash algorithm where the
  84    chain is maintained within the hash table.  The table occupancy is
  85    kept below 75% but we'll assume its perfect:
  86
  87    4 (slot) x 4/3 (occupancy) +  1 (length) = 6 1/3 bytes
  88
  89    So a perfect hashtab has just slightly larger than an average
  90    bcache.
  91
  92    It turns out that an average hashtab is far worse.  Two things
  93    hurt:
  94
  95    - Hashtab's occupancy is more like 50% (it ranges between 38% and
  96    75%) giving a per slot cost of 4x2 vs 4x4/3.
  97
  98    - the string structure needs to be aligned to 8 bytes which for
  99    hashtab wastes 7 bytes, while for bcache wastes only 3.
 100
 101    This gives:
 102
 103    hashtab: 4 x 2 + 1 + 7 = 16 bytes
 104
 105    bcache 4 / 4 + 1 + 4 + 3 = 9 bytes
 106
 107    The numbers of GDB debugging GDB support this.  ~40% vs ~70% overhead.
 108
 109
 110    Speed of bcache VS hashtab (the half hash hack):
 111
 112    While hashtab has a typical chain length of 1, bcache has a chain
 113    length of round 4.  This means that the bcache will require
 114    something like double the number of compares after that initial
 115    hash.  In both cases the comparison takes the form:
 116
 117    a.length == b.length && memcmp (a.data, b.data, a.length) == 0
 118
 119    That is lengths are checked before doing the memcmp.
 120
 121    For GDB debugging GDB, it turned out that all lengths were 24 bytes
 122    (no C++ so only psymbols were cached) and hence, all compares
 123    required a call to memcmp.  As a hack, two bytes of padding
 124    (mentioned above) are used to store the upper 16 bits of the
 125    string's hash value and then that is used in the comparison vis:
 126
 127    a.half_hash == b.half_hash && a.length == b.length && memcmp
 128    (a.data, b.data, a.length)
 129
 130    The numbers from GDB debugging GDB show this to be a remarkable
 131    100% effective (only necessary length and memcmp tests being
 132    performed).
 133
 134    Mind you, looking at the wall clock, the same GDB debugging GDB
 135    showed only marginal speed up (0.780 vs 0.773s).  Seems GDB is too
 136    busy doing something else :-(
 137
 138 */
 139
 140
 141 struct bcache;
 142
 143 /* Find a copy of the LENGTH bytes at ADDR in BCACHE.  If BCACHE has
 144    never seen those bytes before, add a copy of them to BCACHE.  In
 145    either case, return a pointer to BCACHE's copy of that string.
 146    Since the cached value is ment to be read-only, return a const
 147    buffer.  */
 148 extern const void *bcache (const void *addr, int length,
 149                            struct bcache *bcache);
 150
 151 /* Like bcache, but if ADDED is not NULL, set *ADDED to true if the
 152    bytes were newly added to the cache, or to false if the bytes were
 153    found in the cache.  */
 154 extern const void *bcache_full (const void *addr, int length,
 155                                 struct bcache *bcache, int *added);
 156
 157 /* Free all the storage used by BCACHE.  */
 158 extern void bcache_xfree (struct bcache *bcache);
 159
 160 /* Create a new bcache object.  */
 161 extern struct bcache *bcache_xmalloc (
 162     unsigned long (*hash_function)(const void *, int length),
 163     int (*compare_function)(const void *, const void *, int length));
 164
 165 /* Print statistics on BCACHE's memory usage and efficacity at
 166    eliminating duplication.  TYPE should be a string describing the
 167    kind of data BCACHE holds.  Statistics are printed using
 168    `printf_filtered' and its ilk.  */
 169 extern void print_bcache_statistics (struct bcache *bcache, char *type);
 170 extern int bcache_memory_used (struct bcache *bcache);
 171
 172 /* The hash functions */
 173 extern unsigned long hash(const void *addr, int length);
 174 extern unsigned long hash_continue (const void *addr, int length,
 175                                     unsigned long h);
 176
 177 #endif /* BCACHE_H */